> This post is collected in the column *From a Coin to Error Correction* (column index): https://zhichai.net/topic/178636834
Quantum Computing for Beginners: The Math Was Never the Hard Part — It's a Vocabulary You've Never Seen
People often ask: what's the fastest way to build a mental model of quantum computing?
There's a hidden premise in that question: what actually blocks most beginners is not the math. Quantum computing only requires linear algebra — vectors, matrix multiplication, and complex numbers — and anyone who has taken college calculus can go the whole distance. The real blocker is the vocabulary. Superposition, entanglement, measurement, amplitude — you recognize every word, but put together they mean nothing to you. That's not your fault; it's jargon physicists have built up over a century, never intended for engineers.
So a practical strategy: conquer the vocabulary first, and no material will intimidate you afterward. The fastest way isn't grinding through textbooks — it's asking an AI about each concept: "explain superposition using linear algebra," "what operation does measurement actually perform." Restating the jargon in math you already know turns it into a friend. The prerequisite: you need to know which words to ask about.
Below is a vetted resource list in learning order, with notes on why each resource is good, what to look at, and when.
Three Words to Rule Them All
Superposition: a qubit is not 0 or 1, but a complex vector α|0⟩+β|1⟩, where the squared magnitudes of the two components sum to 1. Think of it as betting on 0 and 1 simultaneously, each with a probability weight. n qubits can bet on 2ⁿ combinations at once — 50 qubits means a quadrillion. Sounds like free parallelism, but a fatal trap awaits in the third word.
Entanglement: multiple qubits can be bound into an inseparable whole. Measure one and the other's state is instantly determined, no matter the distance. This is the raw material of quantum computing's exponential expressive power — without entanglement, n bits are just n independent dice, unable to form a 2ⁿ-dimensional space.
Measurement: all the parallelism accumulated above collapses, at the moment of measurement, into a single classical 0 or 1. You only get one answer. The art of quantum algorithms is designing interference structures so wrong answers' probability amplitudes cancel out and the right answer's amplitudes add up — Shor's factoring and Grover's search are both variations on this theme.
Together these three words give you the entire intuition of quantum computing: the space is exponentially large, but there is only one exit. Quantum speedup means using interference to squeeze the structure hidden in that exponential space into a single measurement. Hence a quantum computer is not a "faster computer" — it's only fast on problems with such interference structure. Which problems truly benefit? That leads to the first must-bookmark site.
Learning Resources, In This Order
The Quantum Algorithm Zoo (quantumalgorithmzoo.org). The most worth bookmarking first. It catalogs all known quantum algorithms by category — algebra and number theory, optimization and machine learning, approximation and simulation — each annotated with its speedup over the best classical algorithms, many linking to experimental demos on various platforms. Its value is calibrating expectations: quantum speedup is determined by problem structure, not by buying a new machine. The BQP class it anchors — problems quantum computers solve efficiently and classical machines only polynomially slower — is the discipline's horizon: inside the horizon is industry, outside is physics.
IBM's official tutorials (quantum.cloud.ibm.com/docs/en/tutorials). Free, official, tied to Qiskit — IBM's open-source quantum programming framework, roughly the PyTorch of the quantum world. The unique advantage: a few lines of Python can run your circuit on IBM's actual quantum hardware — not a simulator. Seeing your program produce noisy results from inside a dilution refrigerator feels completely different from reading a hundred pages.
Two textbooks, as needed. *Quantum Computation and Quantum Information* (Nielsen & Chuang) is the international standard — thick, deep, best kept as a long-term reference. Chinese readers now have a better entry option: *Ten Lectures on Quantum Computing*, edited by Sun Xiaoming (researcher at the Institute of Computing Technology, Chinese Academy of Sciences; chair of the CCF quantum computing committee), with deputy editors Shang Yun and Li Lüzhou, published by China Machine Press in March 2024. Nine domestic experts each write the lecture they know best, from fundamentals to algorithms to applications — 336 pages, 89 yuan, the most efficient way for native Chinese speakers to build a conceptual skeleton.
Hardware spec pages. If you're curious about hardware, go straight to IBM Quantum's hardware pages for chip parameters: qubit counts, single/two-qubit gate fidelities, coherence times, error-correction codes. The point isn't memorizing numbers — it's demystification. Every chip honestly lists its error rates; today's quantum machines are "precision instruments that err at every step," still engineering generations away from home appliances.
GitHub summer-camp and hackathon projects from various organizations are also worth browsing: that's the field's real entry-level bar.
The Industry Table, October 2026
Once concepts are built, look at the table. The other half of this list's weight is on the industry side.
Two international poles. IBM's roadmap has reached a critical stage: the 2026 Nighthawk processor targets circuits of 7,500 gates; Starling — the planned first large-scale fault-tolerant quantum computer — is due in 2029, and the modular cryogenic system was first linked up in August 2026. Google, meanwhile, has a Willow hat-trick: in December 2024 it demonstrated that more qubits can mean lower error rates (error correction's lifeline); in October 2025 it ran the first verifiable quantum-advantage algorithm on Willow, 13,000× faster than the fastest supercomputer, published in *Nature*; and just this week came magic-state cultivation news — error rates cut roughly 40×, a new answer to the error-correction resource bottleneck. Both companies share one direction: fault tolerance. Whoever suppresses errors first turns the exponential space into programmable reality.
The Chinese camp. On hardware, Origin Quantum launched its fourth-generation homegrown superconducting machine "Origin Wukong-180" this May: 180 directly operable qubits on a single chip, 99.9% single-qubit gate fidelity. USTC's Zuchongzhi 3.0 beats the fastest supercomputer at random circuit sampling by 15 orders of magnitude, exceeding Google's 2024 result by 6 orders; photonic Jiuzhang 4.0 runs on a different technological route. Domestic software stacks are filling in: Zhongke Quantum Hub (originating from ICT, CAS, with Sun Xiaoming as chief scientist), founded in March, closed two angel rounds within four months, focused on the "Tianshu" quantum operating system and quantum AI error-correction models; UnitaryQuantum released the full-stack open-source algorithm library platform UnitaryLab in May (finance, materials, simulation scenarios) and raised tens of millions of yuan in July. QuantumCTek remains the veteran leader in quantum secure communication. For cloud platform selection, Origin, QuantumCTek, and UnitaryQuantum each occupy a niche: direct hardware access, communication security, algorithm software.
Sandbox AQ deserves a separate look. Spun out of Alphabet in 2022 with over $950 million raised, it builds no hardware — it's the "AI + quantum" software layer, using large quantitative models for numerically intensive problems, with post-quantum cryptography as the other leg (a security business ahead of the quantum computer's arrival). The lesson from this sample: the lightest position in the quantum value chain may not be building machines, but the middleware that connects classical AI to future quantum compute.
As for industry-explainer videos on Bilibili and YouTube, treat them as maps, not textbooks — build a "who is doing what" framework, then verify against the hard resources above.
The Shortest Path
If you have only three weekends: weekend one, talk with an AI about superposition, entanglement, and measurement until you can restate them in your own words; weekend two, work through the first three lectures of *Ten Lectures on Quantum Computing* and run your first Bell-state circuit with IBM's tutorials; weekend three, browse the Algorithm Zoo, pick three algorithms and understand the source of their speedups, then check IBM's hardware pages to see where real machines actually are.
After three weekends, that vocabulary stops being a threshold and becomes a staircase.
---
*References: quantumalgorithmzoo.org; IBM Quantum roadmap; Sun Xiaoming (ed.), *Ten Lectures on Quantum Computing* (China Machine Press, 2024); Nielsen & Chuang, *Quantum Computation and Quantum Information*; Xinhua/People's Daily coverage of Origin Wukong-180 (2026-05);投中网/投资界 coverage of Zhongke Quantum Hub and UnitaryQuantum financing (2026-07); Google Quantum AI Willow results (Nature, 2024–2025).*